Output transformer efficiency - a question.

[Argus25]

Originally Posted by kalee20

"Symmetric clipping implies odd harmonics only"

I'm not 100% certain of this because the anode current approach to saturation and tube cut off are not mirror images/shapes. In the methodology used to calculate the operating conditions by M & H, the tube is not pushed into a clipping condition, so I'm not sure if the calculated bias point and load resistance figures to satisfy B2= 0 would converge on the same values versus the symmetrical clipping case when the tube was over driven.

I wonder if it would be ok with the moderators for me to scan the two pages from the text, citing its origin on the scan and post it ?

[G8HQP Dave]

Symmetric clipping does imply odd harmonics only. Clipping in an SE valve stage will not be symmetric.

Something to watch out for in older books and articles is the idea of balancing off two types of compressive distortion. This certainly reduces second-order distortion, but almost always at the expense of raising higher order distortion - which can sound much worse. Perhaps the worst example of this is using grid current loading of the previous stage to balance off against normal anode nonlinearity in triode voltage amplifiers. You can get impressive dips in second-order distortion, but the rise in 4th 5th etc. will sound awful. This was not fully appreciated 60 years ago. It may be (I don't know) that some of the optimising pentode outputs has some of the same flavour.

[kalee20]

Originally Posted by G8HQP Dave

Symmetric clipping does imply odd harmonics only. Clipping in an SE valve stage will not be symmetric.

I agree - my previous post assumed 'ideal' valves, and with optimum load the positive and negative peaks start to clip at the same level. So distortion is zero up to maximum power, and then ramps up - odd order only.

At other than optimum load, distortion is zero up to a lesser power level, and then ramps up, odd+even order.

As Argus25 points out, the positive and the negative clipping is due to quite different mechanisms, so with practical SE valves, the clipping will not be symmetric anyway. But it can be got fairly close.

[PJL]

There needs to be differentiation between a PP output stage used for audio and one for RF power. This simple article on classes is worth a read http://www.duncanamps.com/technical/ampclasses.html and illustrates how classes other than A describe a range of conduction cycles where class B is a theoretical mid-point of 180deg. You can not build a true class B valve amplifier.

Most audio amps are setup for maximum DC anode current based on anode dissipation and a choice of anode load that leaves it operating largely in class A mode. In class A mode, the power dissipated in the valve reduces as the AC signal increases and part of the constant power is transferred to the speaker.

As the volume is increased, the peaks of our music signal will cause one of the valves to cut-off and we will be operating in class AB. This will increase the total power consumption and if bypassed auto-bias it will change the bias conditions. I would not expect the anode dissipation to exceed the static dissipation.

Increase the volume further and the peaks will start to clip, still further and in the extreme case you have a square wave. The power consumption is now at its peak and theoretically all of the power is transferred to the speaker but as it contains high harmonics, the output transformer has winding capacitance, and the speaker is not a resistance this is not the case and high frequency oscillation will kill an amplifier.

If the signal is further increased, the drive signal will be rectified on the grid and this will lower the bias conditions. Capacitor coupling prevents the amplifier entering AB2 mode. A typical audio PP amp would only do this under fault conditions.

There are various formulae related to constant amplitude sine wave signals which would be of use if designing an RF power amplifier but I don't think they are of much help in analysing audio amps.

[ukcol]

Firstly let me apologies for posting a thread with a question and then largely not joining in. My miserable excuse is that I have had the dreaded "man flu" and my brain has been even duller than usual, making following the replies difficult.

Having more or less recovered and read your posts and the attachments I now have a much better understanding of the audio output stage. One particular aspect that I have got wrong for the past 50 years is the class C arrangement. The description of class B given fits exactly what I always took to be a description of class C, namely I had thought that in class C the valves were biased to almost cut off and not past cut off as is the case.

The classes A, B & C and those in-between can be applied to transistors of course, both discreet transistors and those within integrated circuits.

There is a fourth class of output stage, class D, which as far as I understand is not applicable to valve technology. In class D the output stage is switched between saturation and cut off by a Pulse Width Modulated waveform. The output stage is followed by a low pass filter to remove the switching waveform leaving only the modulating signal. The only example of class D that I can remember coming across was the TDA2600 field timebase IC as used in the Philips G11 chassis. The G11 dates from the 1970s so there must be many examples of class D in use by now. I mention class D because it is an example of very high efficiency; the power wasted as heat in the output device is kept very low.

Now I'm taking my own thread off topic, sorry.

[dseymo1]

Hmm ... is there any reason why Class D couldn't be implemented with valves? The o/p tx could double as the filter, but would perhaps reduce the efficiency somewhat.

[ukcol]

Class D probably could be implemented in valve technology but it would have a high component count, and use a lot of valves. As a result you would lose its main advantage, power efficiency.

[kalee20]

Originally Posted by ukcol

Firstly let me apologies for posting a thread with a question and then largely not joining in. My miserable excuse is that I have had the dreaded "man flu"

Tough luck - hope you are better now!

Originally Posted by ukcol

There is a fourth class of output stage, class D, which as far as I understand is not applicable to valve technology. In class D the output stage is switched between saturation and cut off by a Pulse Width Modulated waveform...

There are other classes too. Yes, Class B is biased to cut-off (with ideal devices), you have two devices to handle positive and negative portions of the waveform each.

Class C has the output device handling less than half a cycle, a tuned circuit fills in the missing bits by flywheel action. So it's only useful for a (nearly) fixed frequency, such as RF.

Class D is purely on/off as you say. So very efficient as there's no loss mechanism in ideal devices. You can switch a a high frequency and create lower frequencies by PWM techniques, again as you say, so you can make a Class D audio amplifier. And there's no fundamental reason why you can't use valves, except that valves don't make particularly good switches.

Class E is also switched, but unlike Class D where square waveforms are seen, in Class E a tuned circuitcauses switching to happen at zero voltage across the output device (or zero current through it), by carefully having the switching happen at the right instant. It's tricky to get right, but it generates less interference than Class D, and, with 'real devices that do not switch completely instantaneously, it gives lower switching losses.

Other classes do clever things with the supply rails, Class G is basically Class B (or AB) with a low supply voltage - but as the signal gets larger other devices switch to a higher voltage. So the high voltage supply only supplies current when it needs to, and is otherwise unloaded. The switching oif supplies, done with transistors and diodes, can be done seamlessly at different points of the AC waveform. It adds complexity but improves efficiency.

All these are about amplifier efficiency, rather than output transformer efficiency, but the output transformer, as has been pointed out, can be made with efficiencies approaching 100%. The big compromises in design occur when the thing has to pass DC, though even then, losses can be made arbitrarily low...

[G6Tanuki]

Originally Posted by PJL

If the signal is further increased, the drive signal will be rectified on the grid and this will lower the bias conditions. Capacitor coupling prevents the amplifier entering AB2 mode. A typical audio PP amp would only do this under fault conditions..

Lots of P-P audio-amps of my acquaintance had a transformer as the driver, and so could happily be pushed into significant grid-current.

Of course this meant the driver stage had to be designed as a 'power' stage not just a signal-level one - if you're running Class-B push-pull output your driver needs to be low-impedance to avoid distortion when the output-stage draws grid-current.

I used the cute little 6N7

http://www.r-type.org/exhib/aaa0902.htm

as a push-pull driver into a transformer, for a pair of zero-bias 807s [750V on the 807 anodes] in some guitar-amps.

This doesn't just apply to high-power stuff: historically, class-B or "Quiescent push-pull" was a popular way to reduce battery-drain in 1930s portable radios.

http://www.r-type.org/addtext/add019.htm

[Argus25]

Lots of P-P audio-amps of my acquaintance had a transformer as the driver, and so could happily be pushed into significant grid-current.

Of course this meant the driver stage had to be designed as a 'power' stage not just a signal-level one - if you're running Class-B push-pull output your driver needs to be low-impedance to avoid distortion when the output-stage draws grid-current.

Yes that is right, this is explained in the article on the UX-171 amplifier I posted recently on this thread. Because of this, driver transformers for push pull class B are step down transformers and the Class A driver stage needs to deliver power to supply the grid current for the class B output stage. On the other hand, class A or AB1 push pull, the driver transformer is a step up transformer and negligible power output is required from the class A driver as no grid current flows.

One very interesting tube made by RCA for class B operation is the 49. This tube is made specifically for class B. With zero bias there is a very low anode current. When the grid is driven positive there is grid current. It avoids bias batteries. Other tubes designed for class A , like the UX171, need a large bias of -50V or more to get them into a class AB2 or class B condition.

[PJL]

Originally Posted by G6Tanuki

Lots of P-P audio-amps of my acquaintance had a transformer as the driver, and so could happily be pushed into significant grid-current.

Agreed, driver transformers are very common in 6V6 16mm projector amplifiers although I expect the reason is not to drive the output valves into grid current but to increase gain.

Audio valves for the most part have been designed to operate in class AB without grid current as illustrated in their data sheets which rarely show performance under +ve grid voltage conditions.

The battery valves and 6N7 are exceptions where the valve designer has been given quiescent power constraints and has produced a valve intended for close to true class B operation. Kind of audio with heavy compromises...and the 807 doesn't count at all .

PS: I just noticed that Argus25 also mentioned the step-up purpose of some driver transformers.